Professional food photographer vs. artificial intelligence (a.i.) to see who is better
Introduction
Artificial Intelligence; it’s been on everyone’s minds lately. It’s been all over our social media feeds. And it may be the demise of us all. So that’s fun. I’ve been asking myself lately: “Is AI really that much better than a professional photographer?” So, in the pursuit of that answer, I took an unexpected creative journey to see who prevailed: humans or AI.
What is A.I.?
So what is AI? Ai is “defined as a machine's ability to perform a task that would've previously required human intelligence.” (1) When we talk about AI, there are a lot of different types, but for this video we’re going to focus on one in particular, and that is: generative AI.
Generative AI is what it sounds like, where it can create a wide variety of data, such as images, videos, audio, text, and 3D models. It does this by learning patterns from existing data, then using this knowledge to generate new and unique outputs (2). Basically, you write a command and the program will produce a piece of creative work.
Programs like Dall-E-2, Midjourney, and Adobe Firefly are all examples of this type of AI, and naturally, many creatives are terrified of this particular AI development, as it may have the potential to take jobs or entire industries away. So, super cute. Love that for us.
But, as a photographer who has been in this industry for nearly a decade, I wanted to explore this AI for myself and see if it could really produce like the pros, or at least, human.
Using AI to craft 3 photo prompts and the different tools
So, my plan is very simple. In order to best compare not only AI to a human photographer, but to see how each AI tool performed against each other, I needed to come up with specific photo prompts that both I and the AI would use to create our images. For this, I’m using a program you might have heard of called ChatGPT. Which is a language-based AI tool that can answer various questions, assist in tasks, and basically help write anything you want.
I started off first by giving the AI some general overview of what I’m trying to do and what questions it had for me in order to write a stronger prompt. It didn’t take too long to answer these along with giving it some added details about what I envisioned for each image.
My goal is to create three different prompts. The first one must feature a single layer chocolate cake. I knew I wanted it to have silky smooth chocolate icing and chocolate bar shavings done into curls on top, and maybe some cocoa nibs for texture. It wanted a chocolatey explosion.
The second must feature a drink, so I’m going with an iced latte. For the latte, I was inspired by this Tik Tok video where this gal made a tiramisu drink and it looked so, so delicious.
And the third and final one is going to be a summer salad. I came across this Instagram account where she creates all these incredible salads. I mean, come on, these all look stunning! How could I not use this as inspiration? It would be a crime not too. I wanted to choose a salad with approachable ingredients but complicated enough to potentially throw the AI off my tracks. I was being a sneaky little minx.
So, after some time chit chatting with ChatGPT and tweaking some notes a bit, I finalized the three image prompts. And, I have to say, I’m really happy and excited to see what we create. I wanted something that was a challenge, and I think we have it.
Baking an AI recipe
Before I dove into taking my photos, I first had to make the cake. Now, for this recipe, I saw this Tik Tok where this guy had AI write him a Pepsi cake recipe, so I had to try it and give it my full review.
Reading through the ingredients, it seems very similar to a southern style coca-cola cake but instead pepsi. I’m a little nervous because it says to make the cake batter using a blender. So, that’s a confusing start.
The verdict? After attempting to bake this cake, I can 100% confirm that this AI generated recipe was trash. Words can not describe how horrible this cake was. It rose up but then collapsed in on itself due to the lack of flour. The texture was like a really squishy pancake and it smelled like fish, not a great description for CAKE. Luckily, I expected this type of outcome to happen, so I made a chocolate cake from King Arthur Baking instead.
My thoughts on the implications of AI
Before I share the images, I want to talk a bit about AI and my thoughts on its implications. My first initial feelings when it comes to AI are, like all of us: I’m scared. I’m fearful for what the future holds when it comes to AI. How will it be used? Will it truly take away jobs, not only for myself, but for the many other creatives in a variety of industries? Is all AI bad? Is any of it good? In the words of Katy Perry, will we all be chained to the rhythm?! Slaves to the machine?
It leaves me asking a lot of questions with little to no security in the answers.
Like many creatives, we have spent years building up our craft and our businesses and to have these pieces of technology come along and potentially sabotage that is both frightening and frustrating. It makes us feel powerless to the inevitable.
I’ve been seeing a lot of photographers not take the growing threat of AI as seriously as I think they should be, because generative AI is technology that is designed to always be learning and improving. And that the longer it spends learning and improving, the better and more accurate the art it creates will become.
With all this doom and gloom, there are some glimmers of hope to consider as we talk about the future of AI. AI can take over some tasks online, but authenticity and ethicality are huge issues for brands (3). Along with who really holds the copyright to these generated images?
In my opinion, I think we’re going to see a lot more conversations around the copyright and ethics side of AI and how it comes more into play as AI progresses, along with potential legislation to protect creatives against loss jobs.
But, there are really amazing AI tools out there that we as creatives can utilize to save time and money. It’s something that can help us, but potentially harm us too. So it really is a double edged sword now isn’t it?
My only hope is that we as humans stand up for each other and fight and crave human crafted art. That’s the only way we’re going to survive and thrive. But maybe all these fears are for not if the AI can’t create better art than a human being…
One thing I learned about each program is that both Dall-e-2 and Midjourney both cost money. The only free one, it seems, was Adobe Firefly in beta. Also, I learned that the prompts I had ChatGPT help me write were WAY too long for any of the programs, so I personally had to go back and condense them. What?! All that energy just for me to have to craft the prompts?? That's so counterintuitive.
Finally, I was working with Adobe Firefly, I got this pop up for Content Credentials will be included. Which, upon further inspection, basically says that all images created with AI as metadata, essentially giving full credit for the artwork to the AI. Which I found interesting.
Comparing My Images to A.I.
Here we are! With the AI finally done creating their images, let's get ready to compare. For each set of images, I'm going to first include the the exact same prompt that all of the AI programs and I received in order to create our images. The first is for the single layer chocolate cake.
'Photo of a bright and colorful scene featuring a single layer chocolate cake made in a 9-inch round cake pan. The frosting should be adorned with colorful sprinkles. Take an angled shot of the scene, capturing it from a 45-degree angle. Use hard/dramatic lighting and include various props around the cake. Create a bright and light atmosphere, and the background is yellow.'
First, let’s take a look at my image for the single layer chocolate cake.
These are some additional images I took of the cake before I cut into it and then a slice on the plate, just as variety.
And now let’s see the other three:
Dall-e-2
Photo Credits: Dall-e-2
Midjourney
Photo Credits: Midjourney
Adobe Firefly
Photo Credits: Adobe Firefly
Next, let’s take a look the images for the iced latte. First, the image prompt used:
'Photo of an iced latte in a textured glass on top of a green box against an emerald green background with a layer of cold foam on top and the coffee visible below. Dust a light amount of cocoa powder over the cold foam. Hard/dramatic lighting.'
Now, let's take a look at my images for the iced latte.
And now let’s see the other three:
Dall-e-2
Photo Credits: Dall-e-2
Midjourney
Photo Credits: Midjourney
Adobe Firefly
Photo Credits: Adobe Firefly
Just a quick side note, as I was reviewing each image, I noticed that they were ALL low-quality. I wasn’t able to download any high-resolution versions of the images. For commercial photographers, this becomes an issue if the client is wanting to put the image on a billboard or even in print. Yes, there are AI programs that can upscale low-res images, but they’re not always reliable.
And, frankly, it’s just yet ANOTHER step we have to take with AI that can be alleviated if you just hired a real person. IT’S REALLY NOT THAT HARD. we’re just creating more unnecessary steps. But I’m getting off track.
Finally, let’s take a look the prompt used for the summer salad photos:
'Photo of a summer salad. Background is a gray stone texture, bright and light. Soft lighting. The bowls are tan with gold forks. Include various props around the scene. The salad itself should consist of fresh mozzarella, tomato slices, sliced grapes, sliced nectarine, and a sprinkle of herbs or micro-greens. Drizzle the salad with rice wine vinegar and honey.'
Here is my images for the summer salad.
And now let’s see the other three:
Dall-e-2
Photo Credits: Dall-e-2
Midjourney
Photo Credits: Midjourney
Adobe Firefly
Photo Credits: Adobe Firefly
The Pros and Cons of A.I.
So, what are the pros and cons of AI? Pros, I think if you’re working on a client mood board and need original images that feature the product, it may be a good supplemental tool. Honestly, time saved is the biggest one. It took me about 30 minutes to create all the AI images from start to finish, which is a huge plus.
But at what cost?
Because the cons are, these images aren’t very good. There are major issues with every single one that needs adjustment. All but one program cost actual money, and you basically have to write the photo prompts in such specific terms to get the output you want. It’s so, so time consuming and tedious.
Ultimately, I felt the AI images lacked any feeling or emotion. It was just a cup of coffee or a salad on a plate, nothing more. It lacked human expression. They felt cold. Marketers' goal are to take compelling creative to sell a product or service, and they can’t do that effectively with imagery that fails to elicit any emotion.
Personally, I believe we should always be fighting for the rights and job security of human beings before AI, but that's just me.